Salma M Bahnasawy (1), Hifza Ahmed (2), Markus Zeitlinger (2), Lena E Friberg (1), Elisabet I Nielsen (1)
(1) Department of Pharmacy, Uppsala University, Uppsala, Sweden, (2)Department of Clinical Pharmacology, Medical University of Vienna, Vienna, Austria.
Background: In vitro time-kill models offer a platform for studying antibiotic PKPD, potentially translatable to in vivo effects via PKPD modelling [1,2]. While Mueller–Hinton broth (MHB) is commonly used in these models, it has a limited resemblance to the in vivo environment [3]. In plasma, factors such as plasma proteins may affect the antibiotic PKPD through plasma protein binding (PPB). PKPD modelling may be used to explore how the presence of human plasma affects the PKPD of antibiotics, and how the translation of PKPD from time-kill experiments could be influenced by the PPB.
Aim: Evaluate the bacterial dynamics and PKPD of cefazolin and clindamycin in time-kill experiments across in vitro media spiked with varying content of human plasma.
Methods: This analysis incorporated equilibrium dialysis data on PPB for cefazolin and clindamycin measured at clinically achievable antibiotic concentrations (2.5, 25 mg/L for cefazolin, and 2, 5, 20 mg/L for clindamycin), in buffer with different percentages of human plasma (20%, 70% human plasma). Static time-kill curve (TKC) experiments provided bacterial count data over 24 hours across three media: MHB, MHB spiked with 20% human plasma, and MHB spiked with 70% human plasma. Bacterial strains included Escherichia coli for cefazolin (MIC: 2 mg/L) and Staphylococcus aureus for clindamycin (MIC: 0.125 mg/L). Antibiotic concentrations ranged 0.25-16 mg/L for cefazolin and 0.0156-16 mg/L for clindamycin. The modelling workflow involved; a) developing PPB models based on total and unbound concentration data, b) developing PKPD models based on TKC experiments in MHB, c) applying the developed PKPD and PPB models to plasma-spiked TKC data without re-estimation, d) sequentially testing the need to scale drug effect parameters using a scaling factor (SF) according to;
Parameter_x% plasma = Parameter_MHB × SF
Results: PPB data included 12 observations for cefazolin and 18 for clindamycin. Bacterial count observations in TKC experiments were 225 for cefazolin and 227 for clindamycin. A linear model described cefazolin’s PPB, with an expected TKCs unbound concentration range of 0.1 – 7 mg/L (20% plasma), and 0.04 – 2.7 mg/L (70% plasma). Clindamycin’s PPB was best described by a second-order polynomial model, resulting in TKCs unbound concentration range of 0.0022 – 6.27 mg/L (20% plasma), and 0.0004 – 2.86 mg/L (70% plasma). Final PKPD models for each antibiotic considered the bacteria to be in growing drug-susceptible or resting insusceptible populations. The drug effect, based on MHB data, was captured by an Emax model (cefazolin; Emax =7.3 h-1, C50=4.96 mg/L, Hill=1, clindamycin; Emax =1.5 h-1, C50=0.024 mg/L, Hill=1). Bacterial regrowth was observed at 24 hours for the cefazolin TKC with concentrations equal to MIC. This was captured in the model as adaptive resistance allowing Emax to decrease with exposure. Bacterial growth lag was observed with the plasma-spiked MHB, but not pure MHB (cefazolin; 0.3 h (20% plasma), 2.7 h (70% plasma), clindamycin; 0.6 h (20% plasma, 1.5 h (70% plasma). Applying the developed PKPD and PPB models to plasma-spiked MHB data was insufficient to describe the observed bacterial growth and killing, necessitating parameter scaling. For cefazolin, C50 was scaled with an estimated SF of 0.72 (20% plasma) and 0.21 (70% plasma). Clindamycin PD was best described by scaling the sigmoidicity factor (Hill) with an estimated SF of 0.24 (20% plasma) and 0.22 (70% plasma).
Conclusions: The developed PKPD models characterised a bacterial growth delay and changes in the antibiotic PKPD in time-kill experiments with plasma-spiked MHB compared to pure MHB when considering only unbound drug concentration to be effective. The need for scaling the drug effect parameters in the presence of plasma could indicate unobserved underlying mechanisms such as nutrient deprivation, bacterial phenotypic changes, or complement system activation. The translation of in vitro time-kill experiment data to in vivo conditions may consequently require considerations beyond measuring unbound drug concentrations. Further experimental research using spiked media could enhance the understanding and prediction of drug activity in vivo.
References:
[1] van Os W and Zeitlinger M. Antibiotics 2021;10:1485.
[2] Nielsen EI and Friberg LE. Pharmacol Rev 2013;65:1053–90.
[3] Nussbaumer-Pröll A and Zeitlinger M. Pharmaceutics 2020;12:773.
Reference: PAGE 32 (2024) Abstr 10879 [www.page-meeting.org/?abstract=10879]
Poster: Drug/Disease Modelling - Infection